Injuries occur frequently among ballet and contemporary dancers. However, limited literature exists on injuries to pre-professional dancers in the USA. The goals of this study were to 1. provide a descriptive epidemiology of the incidence of musculoskeletal injuries in an adolescent and young adult dance population and 2. identify parsimonious regression models that could be potentially used to predict injury incidence. The study was based at the University of North Carolina School of the Arts (UNCSA) from Fall 2009 to Spring 2015. An injury was defined as any event that caused a dancer to be seen at the UNCSA Student Health Services and caused the dancer to modify or curtail dance activity for at least 1 day. Injury rate ratios (IRRs) were calculated using negative binomial generalized estimating equations. Models predicting injury rates were built using forward selection, stratified by sex. Among 480 dancers, 1,014 injuries were sustained. Most injuries were to the lower extremity and the result of overuse. There were differences in upper extremity, lower extremity, and traumatic injury rates by demographic subgroups. Among females, the most parsimonious predictive model for injury rates included a self-reported history of depression, age at time of injury, and number of injuries sustained at UNCSA prior to the semester of current injury. Among males, the most parsimonious model was a univariate model with family history of alcohol or drug problems. Strategies for traumatic injury prevention among dancers should be both sex- and style-specific. No differences were observed in overuse injury rates by sex or style, suggesting that generic overuse prevention strategies may not need to be guided by these factors. It is concluded that strategies can be implemented to reduce and mitigate the consequences of injuries if not the injuries themselves.